The main focus of this work is to study the computational complexity of generalizations of the synchronization problem for deterministic finite automata (DFA). This problem asks for a given DFA, whether there exists a word w that maps each state of the automaton to one state. We call such a word w a synchronizing word. A synchronizing word brings a system from an unknown configuration into a well defined configuration and thereby resets the system. We generalize this problem in four different ways. First, we restrict the set of potential synchronizing words to a fixed regular language associated with the synchronization under regular constraint problem. The motivation here is to control the structure of a synchronizing word so that, for instance, it first brings the system from an operate mode to a reset mode and then finally again into the operate mode. The next generalization concerns the order of states in which a synchronizing word transitions the automaton. Here, a DFA A and a partial order R is given as input and the question is whether there exists a word that synchronizes A and for which the induced state order is consistent with R. Thereby, we study different ways for a word to induce an order on the state set. Then, we change our focus from DFAs to push-down automata and generalize the synchronization problem to push-down automata and in the following work, to visibly push-down automata. Here, a synchronizing word still needs to map each state of the automaton to one state but it further needs to fulfill some constraints on the stack. We study three different types of stack constraints where after reading the synchronizing word, the stacks associated to each run in the automaton must be (1) empty, (2) identical, or (3) can be arbitrary. We observe that the synchronization problem for general push-down automata is undecidable and study restricted sub-classes of push-down automata where the problem becomes decidable. For visibly push-down automata we even obtain efficient algorithms for some settings. The second part of this work studies the intersection non-emptiness problem for DFAs. This problem is related to the problem of whether a given DFA A can be synchronized into a state q as we can see the set of words synchronizing A into q as the intersection of languages accepted by automata obtained by copying A with different initial states and q as their final state. For the intersection non-emptiness problem, we first study the complexity of the, in general PSPACE-complete, problem restricted to subclasses of DFAs associated with the two well known Straubing-Thérien and Cohen-Brzozowski dot-depth hierarchies. Finally, we study the problem whether a given minimal DFA A can be represented as the intersection of a finite set of smaller DFAs such that the language L(A) accepted by A is equal to the intersection of the languages accepted by the smaller DFAs. There, we focus on the subclass of permutation and commutative permutation DFAs and improve known complexity bounds.